We explain the low-energy anomaly reported in several experimental studies of the radiative dipole strength functions in medium-mass nuclei. These strength functions at very low gamma-energies correspond to the gamma-transitions between very close nuclear excited states in the quasicontinuum. In terms of the thermal mean-field, the low-energy enhancement of the strength functions in highly-excited compound nuclei is explained by nucleonic transitions from the thermally unblocked single-quasiparticle states to the single-(quasi)particle continuum. This result is obtained within the finite-temperature quasiparticle random phase approximation in the coordinate space with exact treatment of the single-particle continuum and exactly eliminated spurious translational mode. The case of radiative dipole strength functions at the nuclear excitation energies typical for the thermal neutron capture is illustrated for 94,96,98 Experimental and theoretical studies of the nuclear low-energy electric dipole response remain among the challenges of the modern nuclear structure physics and attract an increasing interest because of its astrophysical impact. Radiative strength (γ-strength) at low energies may enhance the neutron capture rates in the r-process of nucleosynthesis [1, 2] with a considerable influence on elemental abundance distributions. One of the key phases of the r-process nucleosynthesis is capture of a thermal neutron with the subsequent γ-decay of the compound nucleus. The typical neutron energy in the astrophysical plasma is about 100 keV. Therefore, the description of γ-emission spectra of a compound nucleus with excitation energies of the order of the neutron separation energy is the central problem. Hauser-Feshbach model is a standard tool for calculations of the radiative neutron capture cross sections [3]. Formally, this model includes all possible decay channels via transmission coefficients. In the gamma-decay channel the corresponding coefficient is determined by the radiative strength function which is usually calculated by one of the phenomenological parameterizations [4][5][6]. However, in more recent works [1,2,7] it has been shown that for the most important electric dipole strength these simple models are not sufficient because they do not account for structural details of the strength at the neutron threshold. Sensitivity of the stellar reaction rates to these details emphasizes the importance of their studies within microscopic selfconsistent models.Another key ingredient for the Hauser-Feshbach calculations is the Brink-Axel hypothesis [8] stating that the γ-strength does not depend on the nuclear excitation energy, in particular, it is the same for excited and nonexcited nuclei. Supposedly true for the giant resonances and for the soft modes like pygmy dipole resonance, this hypothesis is, however, violated for the lowest transition energies. For instance, non-zero strength is systematically observed at very low gamma-energies [9]. Radiative strength functions extracted from various measurem...